Tariff Tipping Point

125 years of economic data & AI Agents Predict What's Next

Like many economists & data junkies, I've spent most of my professional life believing in the gospel of free trade. It's economic orthodoxy - lower tariffs bring greater prosperity through specialization, competition, and efficient resource allocation. This belief system runs deep, right alongside my passion for data analysis.

When President Trump announced his sweeping "Liberation Day" tariffs on April 2nd, 2025, I felt that familiar anxiety creep in. The markets plunged 2-3% after hours. The baseline 10% tariff on all imports, plus targeted higher rates for specific countries (China at 54%, EU at 20%, Vietnam at 46%), was more extreme than most analysts expected.

Rather than retreat to my pre-existing beliefs, I decided to test them against history. What if the economic theory was wrong, or lacked proper context? Being perpetually riddled with self-doubt (a trait that makes me both a good data scientist and exhausting dinner companion), I turned to what I trust most: historical data and quantitative analysis.

The Analytical Challenge

We have 125 years of well-documented economic history to examine. But I knew my own biases could influence the analysis. So, I did something uncharacteristic - instead of building my own models, I turned to AI agents to handle data collection, processing, and advanced modeling.

I set them loose with a simple directive: find the models that most accurately predict the economic impact of tariffs based on historical data. The results surprised me.

The Historical Context

Let's start with a historical perspective. The data tells an interesting story about how tariffs and economic growth have evolved together:

Trade-Weighted Average Tariff Rates (1900-2025)

• 1900-1920s: 20-30% - Early protectionist policies

• 1930s: ~20% - Smoot-Hawley Tariff Act

• 1950-2000: Decline from ~6% to <2% - GATT/WTO trade liberalization

• 2018-2024: Increase from ~1.6% to ~2.8% - US-China trade tensions

• 2025: 18.8% - Major tariff policy shift

Real GDP Growth & Market Returns (1900-2025)

• 1900-1929: 3.5% - Early industrialization growth

• 1930s: -0.5% - Great Depression contraction

• 1940s: 5.5% - WWII mobilization and post-war boom

• 1950-1970: 4.1% - Post-war economic expansion

• 1970-2000: 3.2% - Mixed growth with recessions

• 2000-2020: 1.8% - Slower growth, financial crisis

• 2025: 0.5% (projected) - Impacted by tariff policy

Meanwhile, the global economic landscape has transformed dramatically:

Global GDP Shares

• 1900: USA (16%), Europe (40%), China (13%), India (8%)

• 1950: USA (40%), Europe (30%), China (4%), India (4%)

• 2000: USA (30%), Europe (25%), China (4%), India (4%)

• 2025: USA (24%), Europe (20%), China (18%), India (8%)

This data shows how America's economic position evolved from challenger to dominant force and now to one major player in a more balanced global economy. The context for tariff policy today is vastly different than it was a century ago.

The Models - Beyond Simple Linear Thinking

This is where things get interesting. My analytical agents didn't select the straightforward models I would have chosen. Instead, they built three sophisticated approaches:

1. Polynomial Regression Model: Captures non-linear, U-shaped relationships between tariffs and economic outcomes.

2. Piecewise Regression Model: Identifies potential "breakpoints" where the relationship between tariffs and growth fundamentally changes.

3. Regime-Switching Model: Recognizes distinct "regimes" where entirely different economic relationships apply.

The regime-switching model proved most accurate in explaining historical patterns - fascinating because it suggests there's no simple linear relationship between tariffs and growth. Instead, the economy operates differently under different tariff environments.

Quintile Analysis - The Sweet Spot

When my agents divided tariff rates into quintiles and analyzed performance, the results were surprising:

The second quintile (3.3%-6.1%) showed the strongest overall economic performance. The fourth quintile (12.5%-18.8%) - which includes our new 2025 tariff rate - showed the weakest performance. Interestingly, the fifth quintile (21.1%-27.5%) performed better than the fourth, suggesting a potential non-linear relationship.

This challenges both extreme positions. Zero tariffs weren't optimal, but neither were very high tariffs.

The Threshold Effect - A Critical Discovery

The regime-switching model identified a critical threshold around 13%. Below this level, economic dynamics follow one pattern; above it, they change dramatically. In the low tariff regime (≤13%), GDP growth actually increases with higher tariffs. But in the high tariff regime (>13%), there's a severe economic penalty followed by recovery as tariffs increase further.

This explains the quintile analysis. The worst performance occurs when crossing this threshold - hence the fourth quintile showing the poorest results.

Before and After 1950 - A Fundamental Shift

Another fascinating insight emerged when my team split the analysis into pre-1950 and post-1950 periods:

Pre-1950 Period:

• Average tariff rate: 16.3%

• Average GDP growth: 2.18%

• Correlation between tariffs and GDP growth: +0.007 (slightly positive)

Post-1950 Period:

• Average tariff rate: 5.69%

• Average GDP growth: 2.67%

• Correlation between tariffs and GDP growth: -0.104 (negative)

This reveals a fundamental shift in the tariff-growth relationship. Before 1950, higher tariffs had a slight positive association with growth. After 1950, as global economic integration accelerated, higher tariffs showed a negative association.

This suggests context matters enormously. What worked in one economic era may not work in another.

Predictions for 2025

What does this mean for our current situation? The models offer the following predictions for the 2025 tariff rate of 18.8%:

The 2025 tariff rate falls within the fourth quintile range (12.5%-18.8%), which historically showed the worst economic performance with an average GDP growth of 0.3% and average market returns of -0.9%.

The regime-switching model - which showed the strongest historical predictive power - offers the most pessimistic outlook, suggesting a potential recession.

Beyond Economic Orthodoxy

This analysis challenges several pieces of economic orthodoxy:

1. Optimal Tariff Range: Consistently across models, moderate tariffs in the 3-6% range appear most advantageous for both GDP growth and market returns.

2. Zero Tariffs Not Optimal: While conventional theory often advocates for free trade, extremely low tariffs (near zero) don't consistently produce the best outcomes. The first quintile (1.4%-2.8%) underperformed the second quintile (3.3%-6.1%).

3. Non-Linear Relationship: All models point to a non-linear relationship between tariffs and economic performance. Simple linear models miss important threshold effects.

4. Historical Context Matters: The relationship between tariffs and growth has evolved. In the modern, globalized economy (post-1950), higher tariffs tend to show a stronger negative association with growth.

Additional Considerations: Change & Uncertainty

Historical data is great at helping us see patterns emerge within the ranges we’ve seen in the past. While the tariff sizes have been seen before over the past 125 years, we have not, historically, seen such a massive change. Even the peak of Smoot Hawley happened over 4 years. The ‘liberation day’ tariffs represent an increase of more than 6x in under 1 year.

Conclusion - Data Over Doctrine

While moderate tariffs may offer some economic advantages, the 2025 tariff levels exceed the range historically associated with optimal economic performance. The regime-switching model indicates that the 2025 tariff rate crosses a critical threshold where economic relationships fundamentally change.

Economic outcomes depend on many factors beyond tariff policy, including global economic conditions, monetary policy, fiscal policy, productivity growth, and geopolitical stability. But based purely on historical patterns, we should expect significant economic challenges ahead.

The data doesn't fully support either extreme position in the tariff debate. Instead, it suggests a more nuanced truth: some level of tariffs may be beneficial, but we've likely crossed the threshold where they become harmful.

In our politically polarized environment, it's tempting to take strong positions. But the data points to a more nuanced middle ground - as it so often does. Perhaps this is the real lesson: economic policy, like good data science, requires continuous questioning of assumptions and a willingness to follow the evidence wherever it leads.

[Note: This analysis focuses solely on economic impacts and doesn't address national security, sovereignty, or other non-economic considerations that may influence tariff policy.

I’m currently working on a follow-up applying game theory to explore potential impact of retaliatory responses. Subscribe to be notified when the next report is ready.

Read the original post and subscribe for updates here.